Background technique
Nerve electric stimulation technology is to pass through electro photoluminescence tune in brain specific region or spinal cord implant electrode using surgical operationThe activity of related Neurons is controlled, to achieve the purpose that treat the nervous system disease.The more traditional damage hand of nerve electric stimulation technologyArt has comparatively safe, the reversible and postoperative advantages such as adjustable, in some nerveous systems such as essential tremor, parkinsonismSignificant curative effect is achieved in system disease.
In recent years, nerve electric stimulation has become the new technical means of clinical intractable epilepsy treatment, wherein vagus nerveElectro photoluminescence (vagus nerve stimulation, VNS) and bilateral thalamus pronucleus electro photoluminescence (Anterior ThalamicNucleus, ANT) treatment intractable epilepsy has been obtained for European Union CE and U.S. FDA authenticates.Both technologies are using openingRing electrical stimulation method applies the periodicity of duration to the neck vagus nerve for the treatment of of intractable epilepsy or bilateral thalamus pronucleusElectro photoluminescence, without considering whether patient itself epileptic attack really occurs.Although this method achieves certain effect, stillThere is also certain problem, i.e., periodical and prolonged electro photoluminescence is difficult to assess on the influence of brain bring.It grindsStudy carefully and point out, inappropriate parameters of electrical stimulation not can be reduced seizure trequency and time not only, or even can induce and deteriorate epilepsyBreaking-out.Compare the electronic stimulation means of open loop, and closed loop electrical stimulation method then needs to carry out the brain electricity of patient prolongedContinuous monitoring gives reactive electro photoluminescence appropriate when patient's brain ammeter reveals epileptic attack feature, realizes the suppression to epilepsySystem, fundamentally reduces the time to the unnecessary electro photoluminescence of patient, so that the possible side effect of electro photoluminescence be minimized.Real-time monitoring and feedback are carried out to the brain electricity after electro photoluminescence in closed loop feedback system simultaneously, electro photoluminescence ginseng can be advanced optimizedNumber.From therapeutic effect, closed loop electro photoluminescence can more effectively reduce epileptic attack number and the breaking-out of epileptic patientDuration, therapeutic effect outline are better than the stimulating method of VNS.
For closed loop electrical stimulation technology, different stimulation target spots, stimulation parameter, stimulation mode are to therapeutic effect and diseasePeople more after also have Different Effects.The starting of epileptic attack has inseparable pass with diffusion and some loops in brain networkSystem can other structures even more extensive region mind in indirect adjustments and controls loop by a certain target spot in electro photoluminescence epilepsy loopExcitability through member is finally reached the purpose for reducing epileptic attack to influence the electrical activity of brain partly or wholly.Therefore rightIn different type epilepsy, the Effective target site on corresponding neural circuitry is selected, would be even more beneficial to precisely controlling to intractable epilepsyIt treats.Detecting the algorithm of epilepsy generation jointly relative to multizone, the independent detection algorithm of multizone, which is more advantageous to, propagates epilepsy,The type of epileptic attack is judged.Only pass through list in most of epilepsy research direction and commercial stimulator at present simultaneouslyA region or multiple regions stimulate after detecting jointly single region, focal zone of the big more options in epilepsyDomain, this mode still need further to be studied for lacking the function and effect of the case in clear epileptic focus region.In addition, insaneThere is also multiclass selection in terms of the stimulation parameter that epilepsy inhibits, pertinent literature has been reported that high and low frequency single channel reaction equation electricityStimulation has the object model of inhibitory effect and its effect and stimulated zone to have direct connection to epilepsy, and for inhomogeneityThe electronic stimulation of type medically intractable epilepsy, pulse width, frequency of stimulation, stimulation time etc. are at present without specificity ginsengNumber.In addition, for epilepsy stimulation mode there is also different stimulus types, multichannel electricity thorn is compared in such as single channel electro photoluminescenceSwash, the more asynchronous multichannel electro photoluminescence etc. of synchronous multichannel electro photoluminescence.It is more that multichannel electrical stimulation pattern provides multizoneThe possibility of road electro photoluminescence, the current strength needed for advantageously reducing on single channel, in stimulation mould compared with single channel electro photoluminescenceThere are more selections on time of formula, spatial distribution, increase the selectivity and validity for the treatment of, be electro photoluminescence as a kind ofThe effective neuromodulation means for treating epileptic condition provide feasibility.
Summary of the invention
In view of the above-mentioned problems, the purpose of the present invention is realize a kind of multichannel closed loop nerve electric stimulation for epilepsy therapySystem, to realize the independent detection to multiple regions epileptic attack situation, to apply in real time to corresponding epileptic attack regionSuitable electro photoluminescence.
The technical solution adopted in the present invention is as follows: a kind of channel wireless radio multi closed loop nerve electric stimulation for epilepsy therapySystem, the system comprises:
Multi-channel signal acquiring module, for the acquisition and analog-to-digital conversion of multiple channel nerve signals, and will be after conversionNerve signal is sent to closed loop control module；
Closed loop control module, closed loop control module are implanted into two-stage series connection classifier, for multi-channel signal acquiring mouldNerve signal after block digital-to-analogue conversion carries out two-stage treatment, and the sieve of doubtful epilepsy nerve signal is carried out in first order classifierChoosing；It screens the signal passed through and enters second level classifier, second level classifier is connected by multiple sub- grade classifiers and formed；SecondCharacteristic value corresponding to the Weak Classifier in sub- grade classifier is first calculated in grade classifier, then by this feature value and the weak typingThe threshold value comparison of device, obtains corresponding output, and what the output and previous sub- grade classifier exported adds up and as the sub- grade classifierOutput, the output of sub- grade classifier is compared to obtain classification results with the sub- grade classifier threshold value；If classification results are notThat doubtful epilepsy signal then stops calculating, and waits the arrival of future time sequence, only classification results be epilepsy signal mindThrough signal, then need to calculate feature corresponding to all Weak Classifiers that all sub- grade classifiers are included, thus independent judgmentWhether each channel occurs epilepsy；
Signal transmission and memory module, each module running parameter and closed loop control module for receiving host computer configuration passDefeated nerve signal, and nerve signal is stored, as object individuation data collection；
Host computer, for being classified according to the two-stage series connection being implanted into object individuation data collection training closed loop control moduleThe parameter of device, and real time communication is carried out with memory module with signal transmission, realize the transmission and data exchange of control instruction；ConfigurationWith running parameter required when adjustment modules work, update in the two-stage series connection classifier being implanted into closed loop control moduleVarious parameters, and collected neuro-physiological signals of real-time display；
Multichannel may be programmed stimulating module, result or host computer instruction for being obtained according to closed loop control module, realWhen change the output of multichannel electro photoluminescence, according to the epilepsy testing result of closed loop control module, to single or multiple intracranial stimulations electricityPole implanted region carries out electro photoluminescence.
Further, the first order classifier in the two-stage series connection classifier is obtained using the training of Ada Boost algorithmStrong classifier.
Further, the sub- concatenated sequence of grade classifier in the second level classifier is weak according to include in the sub- gradeThe validity of classifier is arranged successively from big to small.
Further, the Weak Classifier all passes through the training of Real AdaBoost algorithm combination object individuation dataIt gets.
Further, the two-stage series connection classifier is obtained using the algorithm training of machine learning.
Further, the Weak Classifier in the two-stage series connection classifier is by multiple signal time domains, frequency domain, complexity etc.Feature training in hyperspace obtains.
Further, the feature preferred amplitude, wire length, peak value number, subsegment energy, energy accounting and entropy etc..
Further, the stimulation circuit that the multichannel may be programmed stimulating module is isolated with other circuit electricals.
Also without mentioning closed loop feedback multichannel stimulation circuit
Compared with the existing technology, beneficial effects of the present invention are as follows:
Channel wireless radio multi closed loop nerve acquisition stimulating system provided by the invention is capable of the nerve of independent detection multiple regionsWhether electric signal is epilepsy signal, and suitable parameters of electrical stimulation can be adjusted according to the testing result of epilepsy, and feed back toThe electrical stimulation module of multichannel accurately monitors specific region and neuromodulation so as to realize.This is to deeplyThe neural circuitry that epilepsy is propagated is studied, and applies various combination electrical stimulus patterns in epilepsy propagation loop to optimize electro photoluminescenceIt is of great significance to the inhibitory effect of specific types of epilepsy.
In addition, epilepsy detection algorithm onboard in the present invention can be in conjunction with the think of of Real AdaBoost algorithm and cascadeWant for the various dimensions feature of epilepsy to be combined, realizes that high-performance, the epilepsy of low-power consumption detect algorithm.The algorithm can be generalLow-power consumption microcontroller platform carries out the independent detection of multichannel epilepsy signal, only applies phase in the region for detecting epilepsy signalThe electro photoluminescence answered.This stimulation protocol can reduce unnecessary electro photoluminescence, reduce electro photoluminescence and make to object bring pair is appliedWith.
Specific embodiment
Following further describes the present invention with reference to the drawings.
As shown in Figure 1, the one of specific implementation example of the present invention mainly includes signal acquisition module 1, closed-loop control mouldBlock 2, signal transmission and memory module 3, host computer 4, multichannel may be programmed stimulating module 5.The signal acquisition module 1 canBy connecting implanted electrode such as fibril electrode, ECOG electrode, array electrode, the acquisition nervous physiology telecommunications such as deep brain electrodeNumber, and collected multi-channel nerve analog signal is converted into digital signal, then the nerve signal after conversion is sent toClosed loop control module 2.Signal transmission can pass in wired or wireless manner the nerve signal received with memory module 3Host computer 4 is defeated by so that signal shows and analyzes in real time, and under the operational mode of low-power consumption, signal transmission and memory module3 do not establish physical connection with host computer, directly by the nerve signal received storage into onboard SD memory, for subsequentOff-line analysis and processing.Closed loop and control module 2 are filtered the pre- places such as denoising to data after receiving electroneurographic signalWhether the region of reason, feature extraction and classifying, the single or multiple acquisition channel connections of real-time judge occurs epilepsy, this classification resultsAfter being transferred to stimulation control module, control module is stimulated to configure different electricity from the result of classification according to configured stimulation parameterStimulus modality and parameter simultaneously descend into the programmable stimulating module 5 of multichannel.Multichannel may be programmed stimulating module 5 according to receivingStimulation parameter apply corresponding electric stimulation pulse the discharge scenario of the neuron of the encephalic of corresponding region intervened, completeClosed loop intervention of the system to epileptic attack.
In specific one embodiment of the invention, the signal acquisition module 1 can be received by closed loop control module 2The configuration parameter that host computer 4 passes down completes the initialization of module.Wherein according to the sample rate in configuration parameter, the choosing such as filtering parameterSelect the local field potentials (Local Field Potential) of the intracranial part with different physiological significances, spike potential (Spike)；Reliable letter of the electroneurographic signals such as Cortical ECoG signal (Electrocorticography) as closed loop stimulation systemBreath source.Specifically, the electrophysiological recording integrated chip of the signal acquisition module 1 using miniaturization, chip input terminal it is included everyStraight coupled capacitor can filter out electrode and contact the polarizing voltage generated with brain tissue, therefore can directly be connected with recording electrode.ChipInside include low-noise programmable bandwidth signal amplify array, multiplex analog-digital converter, can acquire EEG, ECoG, LFP,The electricity physiological signals such as Spike, ECG and EMG.Collected analog signal pass through the converter of chip interior output number letterNumber, and the nerve signal after conversion is transferred in closed loop control module 2 by SPI communication mode.
In the specific example of the invention, the further structure of the closed loop control module 2 is as shown in Fig. 2, the module is mainBy data reception module 2.1, data preprocessing module 2.2, epilepsy detection module 2.3, parameter configuration module 2.4, stimulation control5 module compositions such as module 2.5.Wherein data reception module 2.1 is as data acquisition module 1, signal transmission and memory module 3With the interface of closed loop control module, can be responsible for receiving the mind transmitted with buffered signal acquisition module 1 by SPI communication modeThrough signal, and the host computer configuration parameter etc. transmitted by signal transmission with memory module 3.When in data reception module 2.1Data buffer zone obtain the nerve signal time series of predetermined length after, which can be used as Signal Pretreatment mouldThe notch filter of data, the pretreatment such as filtering are completed in the input of block 2.2 in signal pre-processing module 2.2.By pretreatedSignal will enter epilepsy detection module 2.3, and the essence of the module is the classifier of two-stage series connection, judge current this section nerve letterIt number whether is epilepsy segment.The neural deta in multiple channels is independently made whether as the judgement of epilepsy signal, each channel classificationAs a result it will be transferred to stimulation control module 2.5, stimulation control module 2.5 will be according to by upper in parameter configuration module 2.4The corresponding configuration parameter such as frequency of machine transmission, pulsewidth, each channel in the boost pulses such as amplitude parameter and epilepsy detection module 2.3Classification results the parameter of different stimulus modalities is conveyed to multichannel stimulating module 5, if there is region corresponding single or multipleThe nerve signal of channel acquisition is judged as epilepsy signal, then the region will will receive corresponding electro photoluminescence.
In the specific example of the invention, signal transmission with memory module 3 can with WiFi wirelessly or two kinds of formulas of USB with it is upperPosition machine 4 is communicated, and uploads collected former electricity physiological signal perhaps treated electricity physiological signal to host computer or receptionThe configuration parameter sent by host computer and instruction etc..Under low-power consumption mode, signal transmission module with host computer without communicating,But transmitted collected electricity physiological signal data in onboard SD card by SPI communication mode, form object individual characterChange data set so that offline data are analyzed.
In the specific example of the invention, the host computer 4 is assembled for training according to the data stored in signal transmission and memory module 3Practise the personalized two-stage series connection classifier suitable for corresponding channel.The quantity in data sampling region according to each object, it is upperMachine can train multiple personalized two-stage series connection classifiers of corresponding number for each object.
USB mode and wireless mode may be selected after starting in a specific embodiment in the host computer 4.In USB modeIt is connected by scanning USB device, in wireless mode, is attached by the IP address for believing distributed.After connection, pass through readingParameter in configuration register checks 1 connection of signal acquisition module and specific running parameter.It simultaneously can be by upperThe input of 4 interface of machine updates signal acquisition module 1, closed loop control module 2, and multichannel may be programmed the running parameter of stimulating module 5.WhenThe configuration of all modules is correct and after initializing successfully, can run the display of host computer 4 and data that record acquires.Host computer 4 can be intoOne step is such as filtered to the data progress Digital Signal Processing of acquisition and feature extraction.Host computer 4 can manually or operation closed loop is anti-Feedback algorithm automatic trigger electro photoluminescence instruction may be programmed stimulating module 5 to multichannel.Host computer 4 is configurable independently to be transported into low-power consumptionThe transmission of row mode, i.e. signal stops uploading data with memory module 3, and Wi-Fi enters suspend mode, carries out data with low sampling rate and adoptsCollection, stimulation and SD card storage.
Further, host computer 4 using wired or wireless equal communication modes include but is not limited to RS232 serial ports, RS485,Physical path is established between USB, Zig-Bee, bluetooth, Wi-Fi, UWB and signal transmission and memory module 3.The invention specific oneHost computer 4 can be led to by wired transmit with Wi-Fi wireless mode with signal of USB with memory module 3 in system in exampleLetter.Configuration parameter adjustable in system such as sample rate, filter bandwidht, stimulation are joined in system debug or initial phaseVarious parameters etc. in number, classifier are stored in other each modules that transmission module 3 is loaded into system by signal.BelievingNumber acquisition when the nerve signal that signal acquiring board 1 transmits is transmitted to host computer.Classifier parameters in the module updateWhen, the good classifier parameters of 4 off-line training of host computer are loaded into control closed loop module 2 by host computer 4.
In a specific example, the multichannel may be programmed stimulating module 5 can real-time reception by parameter configuration module 2.4The stimulation parameter of storage transmitted by host computer 4 obtains in real time with the multichannel epilepsy testing result in stimulation control module 2.5Different stimulated mode parameter, apply suitable electric pulse in corresponding encephalic epileptic attack region in time, before interfering epileptic attackThe dynamic change of phase neuroid inhibits the propagation and breaking-out of epilepsy.
Closed loop control module 2 described in one of them further embodiment is used to be constituted in two-stage series connection classifierWeak Classifier and Weak Classifier waterfall sequence, be by Real AdaBoost algorithm be based on so that in training set it is positive and negativeThe smallest standard of loss function in sample set obtains.Wherein Weak Classifier ciIt is made of threshold value and the output function of segmentation, whenThe corresponding characteristic value f of signal is greater than threshold θ and then exports a numerical value, otherwise exports another numerical value.Weak Classifier outputPiecewise function and threshold value obtained by the training of the collected nerve signal in each channel.
The classifier H (x) of the first order is obtained by the training of Real AdaBoost algorithm.Weak point used in the first orderClass device ciThe corresponding small feature of calculation amount such as amplitude, wire length etc., conducive to the quick screening of doubtful epilepsy signal.
H (x)=∑I=1 ..., ni(2)
The classifier of the second level is to connect to obtain by multiple sub- grade classifiers, the output d of every grade of classifierkIt is that previous stage is defeatedD outk-1C is exported with Weak Classifier corresponding to the same leveljIt is cumulative, particularly, the output of chopped-off head classifier is that this grade is includedThe output of Weak Classifier.The output d of this grade classificationkWith this grade of threshold value rkIt is compared for determining whether the nerve signal is insaneEpilepsy signal.Only this grade, which is determined as epilepsy signal just, will do it the calculating of next stage, otherwise determine that this epilepsy signal is normal mindThrough signal, subsequent calculating is no longer carried out, to reduce the complexity of calculating.Wherein series sequence qkBy being chosen by greedy algorithmChoosing currently to concentrate the maximum Weak Classifier of difference of positive and negative sample to obtain in training sample.
dk=dk-1+cqk(x) (3)
Further, for the building of the two-stage series connection classifier of epilepsy detection and using respectively in host computer 4 and epilepsy inspectionIt surveys in module 2.3 and carries out.As shown in figure 3, the building process of classifier is divided into the offline classifier building of S1, the offline classifier of S2Assessment, the use process of classifier are S3.The algorithm idea of training dataset and AdaBoost of the algorithm based on specific individualTrain multiple personalized classifiers corresponding to different acquisition region, the data set which can be different according to object, fromUsed threshold value in dynamic adjustment classifier, output piecewise function etc., so that the classifier is directed to the different zones of different objectsReach the optimal effect of classification results, improves and need in stimulation system used in medical treatment at present through medical staff's rootThe treatment mode of stimulation parameter is manually adjusted according to the performance situation of patient's state of an illness.
Further, in the offline classifier building of S1, we are by training dataset by object individual character by way of sliding windowChange the nerve signal sample decomposition in data set into suitable length, forms isometric time series (S1.1 sample decomposition), soIt is special that the feature such as time domain for being easy to distinguish epilepsy segment and normal EEG signals in time series is calculated according to existing experience afterwardsSign, frequency domain character, time and frequency domain characteristics, analysis of complexity feature etc. (S1.2 feature extraction).The spy being made of a series of featureInput of the vector as classifier is levied, each feature in vector is all trained to one by Integrated Algorithm RealAdaBoostA Weak Classifier, the Weak Classifier are made of threshold value and piecewise function.Wherein RealAdaBoost is selected according to greedy algorithm and is madeIt obtains the smallest feature of current training sample set loss function and makees current optimal Weak Classifier, and classifier is divided in certain one kindOutput of the correct confidence level of class as the Weak Classifier, wherein the bigger confidence level of the absolute value exported is bigger, due to this weak pointClass device is two classification, so the function of output is two-value piecewise function.After the completion of the wheel Weak Classifier is selected, which willIncrease the specific gravity of current class device error sample so that the emphasis of subsequent Weak Classifier classification on current wrong point of sample simultaneouslyAgain new optimum classifier is selected until the feature in feature vector is all selected and finished.According to Weak Classifier reliability orderAnd the performance indicator of classifier, choose basic Component units (S1.3 spy of the top n Weak Classifier as epilepsy detection classifierSign is chosen and is calculated).It is insane used in epilepsy detection module 2.3 in order to make algorithm be more suitable for the real-time operation of hardware platformEpilepsy classifier is made of two-stage series connection.First order classifier is chosen the small feature of calculation amount and is obtained by RealAdaBoost trainingOne strong classifier can screen a large amount of non-epilepsy signal since computation complexity is low in short-term.Only it is judged as in the first orderThe nerve signal of doubtful epilepsy segment just needs to enter second level classifier and is handled.The multistage classifier of the second level is by moreA Weak Classifier connects to obtain using Cascade thought, the output d of the every height grade of classifierkIt is previous stage output dk-1With bookGrade Weak Classifier cjCorresponding output adds up.The output d of the sub- grade classificationkWith this grade of threshold value rkIt is compared for determiningWhether the nerve signal is epilepsy signal.Only this grade, which is determined as epilepsy signal just, will do it the calculating of next stage, otherwise determineThis epilepsy signal is normal nerve signals, no longer carries out subsequent calculating, waits the arrival of next clock signal, to reduce meterThe complexity of calculation.It can train to obtain multiple corresponding personalized two-stages for multiple pickup area neural detas in each objectSeries connection classifier, the structure of each two-stage series connection classifier is identical, but due to the difference of data set, each two-stage series connection classifierMiddle Weak Classifier, Weak Classifier put in order, the different two-stage series connection for making the personalization of the parameter of sub- grade classifierClassification of the classifier on corresponding region is optimal.
After obtaining personalized two-stage series connection classifier, which is assessed in S2 by test sample collection.It includes positive negative sample that wherein test sample, which is concentrated, and sample label is demarcated by professional person, the positive negative sample that test sample is concentratedIt is random to occur.The signal that test sample is concentrated is processed into the time series with training sample equal length by sliding window.Work as timingSequence pass through two-stage series connection classifier when classifier every level-one will first calculate the corresponding feature of this grade of Weak Classifier and withIts threshold value compares to obtain the output of Weak Classifier and is added up to obtain this grade with the output of previous stage cascade classifierCascade classifier output, the result then determine that this signal is normal compared with the threshold value of cascade classifier, if it is less than threshold valueEEG signals are then continued to execute if it is greater than threshold value and are finished until all cascade Weak Classifiers are all run, then the signal is dividedClass is epileptic attack signal.Tag along sort that each time series obtains and the label artificially demarcated compare, and obtain this pointSuch as sensitivity of evaluation index of the class device in the test sample collection, specificity, detection delay etc..If current cascade classifier is notMeet evaluation index and then return in S1 and select feature again and be trained, until obtaining the classifier for meeting evaluation index.InstituteThe personalized two-stage series connection classifier to be formed is cascaded so that multiple and different features can be obtained for each object.
After host computer completes the off-line training S1 and assessment S2 of classifier, carries out the hardware transplanting of algorithm and examined in epilepsySurvey in module 2.3 and realize real-time epilepsy hardware detection algorithm, the classification process of nerve signal as shown at s 3, in real time by recordEeg signal classification is normal EEG signals and epileptic EEG Signal.
Further, the real-time process flow of two-stage series connection classifier is as shown in Figure 4 in epilepsy detection module 2.3.The first orderClassifier refers to that, by simple feature such as amplitude, the Weak Classifier that the training such as wire length obtains cascades up to form a strong classifier,The output valve of strong classifier be the accumulation of multiple Weak Classifiers and, if signal is not classified as normally in first order classifierEEG signals then continue to execute the cascade classifier of the second level.The characteristics of such classification cascade classifier, is that calculation amount is small sameWhen by multiple features of signal combine improve classification accuracy, be suitble to hardware algorithm operation early period.
Due to the EEG signals of different objects, the differences such as electrode implanted region, different object different zones epileptic attacksThere is also differences for the performance of EEG signals.Different training datasets are established to generate individual character to the EEG signals of different zonesIt is of great advantage for the performance for improving epilepsy detection classifier to change classifier, and the algorithm idea based on machine learning is based on individual characterThe identical algorithm frame of classifier and personalized parameter are more advantageous to algorithm universality and personalized synchronous requirement.This hairIn bright, when being implanted into hardware algorithm for different objects, it is only necessary to personalized classifier parameters be modified, hardware is reducedIt is implanted into complexity.
In a specific example, the Weak Classifier in the two-stage series connection classifier is by time domain, frequency domain, and complexity etc. is moreFeature training on a dimension space obtains, in conjunction with hyperspace signal the normal EEG signals and epilepsy signal the characteristics of, phaseFeature in single dimension can more comprehensively dissect signal, classification accuracy is improved.
In a specific example, the feature in hyperspace is preferably amplitude, wire length, peak value number, subsegment energy, subsegmentEnergy accounting, approximate entropy etc..The calculation amount of these features is not once overall calculation amount is small, the real-time inspection suitable for epilepsy signalIt surveys.In two-stage series connection classifier, feature used in the first order is preferably amplitude, wire length etc., for doubtful epilepsy signalQuickly screening；How sub- level structure in the second level makes complex characteristic such as frequency domain energy, and entropy etc. also can be before smaller calculation amountIt puts, excavates the depth information of signal, improve classification accuracy.
In one of the embodiments the multichannel may be programmed stimulating module 5 mainly with single-chip microcontroller, constant-current circuit andDC/DC circuit composition.The SPI interface that single-chip microcontroller and closed loop control module 2 communicate individually is supplied with battery by light-coupled isolationElectricity realizes electrical isolation with other circuits, and when stimulating module generates stimulated current, electric current does not flow through collection plate, to subtractThe record artefact that pinprick generates.Single-chip microcontroller receives stimulus modality parameter, and the simulation electricity of random waveform can be exported by DACPressure turns electric current (V/C) circuit through overvoltage, and output constant current waveform realizes multichannel stimulation output eventually by multiway analog switch.